问题
I am working on software for a robot, which is normally run on the Raspberry Pi. Let's consider the imports of two files:
motor.py (runs the motors):
from RPi import GPIO as gpio
and client.py (communicates with the server and relays commands to the motors):
from rpi.motor import Motor
Both files are in a directory called rpi, which contains a __init__.py and a __main__.py. The RPi package cannot be installed on non-RPi devices. However, I still want to test the functionality of client.py.
import unittest
from unittest import mock
# Location A
class TestClient(unittest.TestCase):
# Location B
setUp(self):
# Location C
pass
Originally, I tried from rpi.client import Client at LocA, but that failed because it tried to import Motor, and then import GPIO from RPi, which doesn't exist. I also tried mock.patch("rpi.client.Motor") at LocB (including adding mock_motor after self, and imported Client at LocC, but that failed as well. I tried mocking RPi at LocA, too, but it didn't work either.
How do you mock out a library that is not installed on your system?
回答1:
You can use patch.dict() to patch sys.modules and mock RPi module as showed in pointed documentation.
Use follow code at the top of your test module:
>>> from mock import MagicMock, patch
>>> mymodule = MagicMock()
>>> patch.dict("sys.modules", RPi=mymodule).start()
>>> from RPi import GPIO as gpio
>>> gpio
<MagicMock name='mock.GPIO' id='139664555819920'>
>>> import os
>>> os
<module 'os' from '/usr/lib/python2.7/os.pyc'>
In Python3 you have same behavior.
In your specific case use patch.dict is little bit overkill; maybe you aren't interested in patch context and original state recover. So you can simplify it by set sys.modules["RPi"] directly:
>>> from unittest.mock import MagicMock
>>> mymodule = MagicMock()
>>> import sys
>>> sys.modules["RPi"] = mymodule
>>> from RPi import GPIO as gpio
>>> gpio
<MagicMock name='mock.GPIO' id='140511459454648'>
>>> import os
>>> os
<module 'os' from '/usr/lib/python3.4/os.py'>
来源:https://stackoverflow.com/questions/32508349/python-mock-and-libraries-that-are-not-installed